Re: [Quilt-dev] [PATCH] mail-single-patch.diff

[Andreas Gruenbacher]

On Sunday 18 September 2005 16:30, John Vandenberg wrote:
> On 9/18/05, Andreas Gruenbacher <address@hidden> wrote:
> > On Sunday 18 September 2005 05:22, John Vandenberg wrote:
> > > On 9/18/05, Gary V. Vaughan <address@hidden> wrote:
> > > > This patch allows calling quilt mail with a list of patch names
> > > > after the usual arguments, and mails only those patches rather
> > > > than the entire series.  If only a single patch is named then it
> > > > is put into the bady of a single mail (that's what I'm using to
> > > > send this patch)
> > >
> > > Hi Gary,
> > >
> > > This patch makes `mail' the first command to operate on multiple
> > > patches outside of the series.  While I would enjoy being able to send
> > > a single patch, or a sub-series, I don't think quilt should generate
> > > emails that can't be applied on the receiving end.  In other words, in
> > > order to send an adhoc set of patches, `mail' should first verify that
> > > the dependency graph of the selected patches is sane.
> >
> > It can't really do that without actually trying to apply the patches as
> > specified, and even that doesn't guarantee that all hunks get applied in
> > the appropriate place. (The result of the graph command isn't "precise"
> > enough and doesn't account for fuzzy matches, etc.)
>
> I was thinking more along the lines of permitting selection of a
> number of distinct patches to be emailed.  To account for fuzz,
> annotate could be used instead of graph.

*choke* annotate isn't up to this, either. Let's keep things at a reasonable 
level of complexity.

I'd rather not allow more than a consecutive range of patches per mail 
command.

> > How about specifying a sub-series instead, like the diff command allows?
>
> A sub-series starting from the first patch in the series?  That is
> usually where the sub-series of good patches will be.  a sub-series in
> the middle of the series has the same problems as cherry picking.

Not really. Say you send off five patches, the first two get accepted, and the 
third and fourth needs to be adapted. Then sending only thouse should be 
fine.

> > What I'd like to have in addition in the 0th message is the patch series,
> > optionally with the patch subjects, generated from the patches --
> > something like:
> >
> >   one.diff
> >     Add feature
> >
> >   two.diff
> >     Fix bug
> >
> >   three.diff
> >     Remove dead code
>
> This would be nice.
>
> > Next we probably want to allow sending patches as reply to another
> > message. I'm still short of good ideas with this.
>
> How about the these for bad idea's to get the discussion going .. :)
>
> Can we repost patches by recording the message-id? in the patch
> header?  Reposting would not need the same level of smarts, as the
> patches were once verified, and the recipients are less likely to be
> scratching their heads.
>
> For replies to others, pass a --msgid to mail ?

We'd need the message ids of replies to the original patch too, not only the 
ones we've generated ourself.

-- Andreas.